def propagate_pad(op):
# type: (ONNXOperation)->typing.Tuple[typing.List[typing.List[int]], typing.List[str]]
if 'paddings' in op.attribs:
paddings = op.attribs['paddings']
elif 'pads' in op.attribs:
paddings = op.attribs['pads']
else:
assert False
return [
infer.pad(input=op.input.shape, padding=to_nnef_padding(paddings))
], [op.input.dtype]
def pad_image_shape(op):
# type: (Caffe2Operation)->ShapeResult
is_nhwc = op.attribs.get('order', 'NCHW').upper() == 'NHWC'
def expand(list, default):
if is_nhwc:
return [default] + list + [default]
else:
return [default, default] + list
return infer.pad(input=op.inputs[0].shape,
padding=expand(
caffe2_pads_to_nnef_padding(op.attribs['pads']),
(0, 0))), op.inputs[0].dtype
def propagate_pad(op, const_value_by_tensor):
# type: (TFOperation, _ConstValueByTensorT)->typing.Tuple[typing.List[typing.List[int]], typing.List[str]]
input, paddings = op.inputs
paddings = const_value_by_tensor[paddings].tolist() # type: typing.List[typing.Tuple[int, int]]
return [infer.pad(input=input.shape, padding=paddings)], [op.attribs['T']]
def test_pad(self):
self.assertEqual([2, 7, 12], infer.pad([1, 2, 3], [(0, 1), (2, 3), (4, 5)]))
